Multithreading and Actors

1. MultiThreading
& Actors
Diego Pacheco

3. I don't need to worry about Threads. Really? Wrong!

4. Moore Law and the need for Threads

5. Reduce Latency Matters.

6. Multiprocessing vs multithreading

7. Concurrent Vs Parallel

8. How Should I pick Concurrency or Parallelism ?

10. Block vs Non-Blocking IO

12. Let's not paint a “Rose Picture” on Async
❏ Potentially you can have:
❏ Memory Leaks
❏ Race Conditions
❏ Complex state Machines
❏ Callback Hells (Look NodeJS)
❏ Error Handler can be difficult (look RxJava)

13. A Simple Thread in Java

14. Thread issues
❏ No Language has a great abstraction to model TIME.
❏ Deadlocks
❏ Race Conditions
❏ Memory Inconsistency Errors (shared State)
❏ Bugs & Troubleshooting Issues

15. Oopsie! Something went wrong...

16. Oopsie! Something went wrong...

17. Synchronized to rescue...

18. Synchronized to rescue...

19. Atomics to rescue...

20. Atomics to rescue...

21. Executors

22. Executors
❏ Safe if your code
is safe.
❏ Better
Abstraction then
Direct Threads.
❏ Don't forget to
shutdown.

23. Executors
❏ Less worries.
❏ Thread Management
❏ Works with Queues
❏ Your code still need to be
Thread SAFE.
❏ Create ONCE, re-use as
much as you can.
❏ Be Careful with “clone”.

24. Executors

25. Executors

26. Executors

27. Locks

28. Locks

29. Locks

30. Java Concurrent Collections

33. Threads Exercise (with Java >= 8)
Constraints: You cannot use frameworks, only the
“things” you learned in this class. Use Java.
1. Build a concurrent program with prints the time table of 7 (7x1, 7x2,7x3... 7x10) all in parallel and when
it ends print the results in order.
2. Build a concurrent program which calculates the factorial from 1 to 200 in parallel. Print the results at
the end when all calculations are done.
3. Build a concurrent program that count words in a file wich has 10mb. The counting must happen in
parallel as well.
4. Design and code a Worker from sratch, you should have a Queue and manage threads for the user. You
cannot use frameworks on any high level features like Service Executors or Java Thread Pools - create your
own.

34. Actors & Akka

35. Reactive Stream
❏ Standard Async Stream Processing
❏ Non-Blocking back pressure
❏ Equivalent Reactive Stream -> JDK9 java.util.concurrent.Flow
❏ Handle Stream of Data Issues
❏ Resource Consumption (i.g: Flood of Fast Data)
❏ Right Async code in order to Handle multiple CPU Cores
❏ Governance or Stream Data Exchanges (avoid arbitrary
buffering)
https://www.reactive-streams.org/

36. Reactive Manifesto (Value from Non-Blocking + Async)
https://www.reactivemanifesto.org/

37. Actors & Akka
❏ Mature Actor ecosystem based on Erlang / OTP
❏ “Simple” concurrent Distributed Systems
❏ Resilient By Design
❏ Implement Reactive Streams principles
❏ High Performance & Scalability
❏ Up to 50 million msg/sec on a single machine.
❏ Small memory footprint; ~2.5 million actors 1GB.

38. Actors Issues
❏ Actors Code Reuse is complicated.
❏ Not meant for all kinds of problems.
❏ Does not work well on a “PURE” RPC system. Since actors
systems are similar to state machines.
❏ Can be much more complex than regular service development.
❏ Testing is Painful (being improved) but still.

39. Typed Actors with Scala

40. Typed Actors with Scala
$ sbt compile
$ sbt run

41. Messages == Part of The protocol (Actors are the other part)

42. Create the Actor System and Run

43. Typed Actors with Scala

44. Typed Actors with Scala ~ HelloWorld

45. Typed Actors with Scala ~ HelloWorldBot

46. Typed Actors with Scala ~ HelloWorldMain

47. Actors Exercise (with Akka >= 2.6)
Constraints: You cannot use frameworks besides Akka,
you need to code with Scala >= 2.13
1. Build a Chat application Akka: No Ui is needed. You need to have Messages, ChatRooms, People and people
need to be able to be in multiple chat rooms at same time. Chat Rooms has a limit of 10 people.
2. Build a Chat application Akka - part 2: Enhance your chat application and now create an Admin role
where a user could become an Admin and be able to KICK any user from any chat room.
3. Build a Chat application Akka - part 3: Enhance your chat application and now create observability for the
system, you need to have live tracking of: Number of messages, Number of chat rooms, top 3 people who sends
more message, top 3 chat rooms with more messages. Provide a print operation for the users see the metrics.
3. Build a Chat application Akka - part 4: Enhance your chat application and now Expose operations via
REST interface(for this homework you can add more frameworks and libs).

48. MultiThreading
& Actors
Diego Pacheco